Skip to content

Latest commit

 

History

History
76 lines (57 loc) · 3.51 KB

README.md

File metadata and controls

76 lines (57 loc) · 3.51 KB

Introduction

This project aims to implement a multi-threaded version of the PageRank algorithm. The algorithm is used to rank web pages in search engines, and it is based on the idea that a page is important if it is linked by other important pages. The algorithm is iterative and computes the PageRank values of the nodes in a graph. See the report for firther details.

Project Structure

The project is divided into the following files:

  1. main.cpp: The main file that reads the input graph and calls the PageRank function. It takes arguments for the path to a list of directed graph edges and the maximum number of threads to use. It then measures the execution time from 1 to the maximum number of threads and saves the results in a .csv file inside the stats folder.

  2. main_perf.cpp: Another main file specifically created for analysis using the Perf tool. It takes arguments for the path to a list of directed graph edges, the algorithm to use (a or b), and whether it's sequential or parallel (s or p). It simply calls the required PageRank function.

  3. main_scorep.cpp: Another main file specifically created for analysis using the Score-P tool. It takes arguments for the path to a list of directed graph edges and performs parallel PageRank using all available threads.

  4. utility.h and .cpp: A file containing utility functions such as the function to read the input graph and the function to calculate statistics.

  5. graph/graph.h and .cpp: A file containing the graph class that stores a column-wise graph and the functions to calculate PageRank.

  6. graph/graph_by_row.h and .cpp: A file containing the graph class that stores a row-wise graph and the functions to calculate PageRank.

  7. speedup_graphs.py: A Python script that reads the .csv files inside the stats folder and generates speedup graphs.

  8. graph_generator.py: A Python script that generates a list of random directed graph edges with a given number of nodes and edges.

How to Run

Main File

To compile the project, run the following commands:

g++ -std=c++20 main.cpp utility.cpp graph/graph.cpp graph/graph_by_row.cpp -o main -fopenmp -O3

To run the project, use the following command:

./main <path_to_graph_edges> <max_threads>

Example (Run main on p2p_Gnutella31 up to 26 threads):

./main ./graphs/p2p_Gnutella31.txt 26

Perf Tool

To compile and run the project for analysis using the Perf tool, run the following commands:

g++ -std=c++20 main_perf.cpp utility.cpp graph/graph.cpp graph/graph_by_row.cpp -o main_perf -fopenmp -O3

To run the project for analysis using the Perf tool, use the following command:

perf record -g ./main_perf <path_to_graph_edges> <algorithm> <mode>

Example (Run Algorithm 1 Parallel on p2p_Gnutella31):

perf stat -d ./main_perf ./graphs/p2p_Gnutella31.txt a p

Score-P Tool

To compile and run the project for analysis using the Score-P tool (must be installed before), run the following commands:

mkdir scorep
cd scorep
scorep-g++ -std=c++20 ../main_scorep.cpp ../utility.cpp ../graph/graph.cpp ../graph/graph_by_row.cpp -o main_scorep -fopenmp

To run the project for analysis using the Score-P tool, use the following command:

mpirun -np 1 ./main_scorep <path_to_graph_edges>

This will produce a folder named scorep-<timestamp> containing the analysis results ready to be visualized using the Vampir tool.

Example (Run main_scorep on p2p_Gnutella31):

mpirun -np 1 ./main_scorep ../graphs/p2p_Gnutella31.txt